!-------------------------------------- LICENCE BEGIN ------------------------------------
!Environment Canada - Atmospheric Science and Technology License/Disclaimer,
! version 3; Last Modified: May 7, 2008.
!This is free but copyrighted software; you can use/redistribute/modify it under the terms
!of the Environment Canada - Atmospheric Science and Technology License/Disclaimer
!version 3 or (at your option) any later version that should be found at:
!http://collaboration.cmc.ec.gc.ca/science/rpn.comm/license.html
!
!This software is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY;
!without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
!See the above mentioned License/Disclaimer for more details.
!You should have received a copy of the License/Disclaimer along with this software;
!if not, you can write to: EC-RPN COMM Group, 2121 TransCanada, suite 500, Dorval (Quebec),
!CANADA, H9P 1J3; or send e-mail to service.rpn@ec.gc.ca
!-------------------------------------- LICENCE END --------------------------------------
***s/r hzd_solfft - parallel direct sol_8ution of high-order diffusion
* equation with ffft8
#include "model_macros_f.h"
subroutine hzd_solfft (F_sol, F_Rhs_8, F_pri_8 , 46
% F_a_8, F_c_8, F_deltai_8,
% minx1, maxx1, minx2, maxx2, nx1, nx2, nx3, F_pwr,
% minx,maxx,miny,maxy,gnk,gni,nil,njl,nkl,
% F_opsxp0_8, F_opsyp0_8,F_cdiff,F_npex,F_npey)
*
implicit none
*
integer minx1, maxx1, minx2, maxx2 , nx1, nx2, nx3, F_pwr,
$ minx , maxx , miny , maxy , gnk, gni,
$ njl , nkl , nil , F_npex, F_npey
real*8 F_opsxp0_8(*), F_opsyp0_8(*),F_pri_8,
$ F_a_8(1:F_pwr,1:F_pwr,minx2:maxx2,nx3),
$ F_c_8(1:F_pwr,1:F_pwr,minx2:maxx2,nx3),
$ F_deltai_8(1:F_pwr,1:F_pwr,minx2:maxx2,nx3),
$ F_Rhs_8
real F_cdiff, F_sol(minx:maxx,miny:maxy,gnk)
*
*author
* Abdessamad Qaddouri
*
*revision
* v2_10 - Qaddouri A. - initial version
* v3_02 - J. P. Toviessi - remove data overflow bug
* v3_10 - Corbeil & Desgagne & Lee - AIXport+Opti+OpenMP
* v3_11 - Corbeil L. - new RPNCOMM transpose
*
*object
*
*arguments
* Name I/O Description
*----------------------------------------------------------------
* F_sol I/O r.h.s. and result of horizontal diffusion
* F_Rhs_8 I work vector
*
*----------------------------------------------------------------
*
**
#include "ptopo.cdk"
real*8 fdg1_8 ( miny:maxy, minx1:maxx1, gni+F_npex ),
$ fdg2_8 (minx1:maxx1,minx2:maxx2, nx3+F_npey ),
$ dn3_8 (minx1:maxx1,minx2:maxx2,F_pwr,nx3),
$ sol_8 (minx1:maxx1,minx2:maxx2,F_pwr,nx3),
$ fwft_8( miny:maxy ,minx1:maxx1,gni+2+F_npex)
real*8 ZERO_8, b_8, fact_8
parameter( ZERO_8 = 0.0 )
integer o1,o2,i,j,k,jw
integer ki, kkii, ki0, kin, kilon, kitotal
integer pi0,pin
**
* __________________________________________________________________
*
kilon = (maxx2-minx2+1 +Ptopo_npeOpenMP)/Ptopo_npeOpenMP
*
*** F_Rhs_8(minx:maxx,miny:maxy,gnk)
*
* Resolution
*
fact_8=((-1)**F_pwr)*dble(F_cdiff)
call rpn_comm_transpose48 ( F_sol, Minx, Maxx, Gni,1, (Maxy-Miny+1),
% (Maxy-Miny+1),minx1, maxx1, Gnk, fdg1_8, 1,
% fact_8,0.d0)
!$omp parallel private(ki0,kin,pi0,pin)
!$omp do
do i = 1, Gni
do k = 1, Nkl
do j = 1, njl
fdg1_8(j,k,i) = F_opsxp0_8(gni+i)*fdg1_8(j,k,i)
enddo
enddo
do k = 1, Nkl
do j= njl+1,maxy
fwft_8(j,k,i) = ZERO_8
enddo
enddo
*
do k= Nkl+1, maxx1
do j= miny,maxy
fwft_8(j,k,i) = ZERO_8
enddo
enddo
*
do k=1,nkl
do j= miny,0
fwft_8(j,k,i) = ZERO_8
enddo
do j=1,njl
fwft_8(j,k,i) = fdg1_8(j,k,i)
enddo
enddo
enddo
!$omp enddo
!$omp do
do i= 1,gni+2
do k=minx1,0
do j= miny,maxy
fwft_8(j,k,i) = ZERO_8
enddo
enddo
enddo
!$omp enddo
*
* projection ( wfft = x transposed * g )
*
!$omp do
do k=1,Nkl
call ffft8 (fwft_8(miny,k,1),(Maxy-Miny+1)*(maxx1-minx1+1),
% 1, (Maxy-Miny+1) , -1 )
enddo
!$omp enddo
*
! b_8 = F_pri_8
!$omp do
do i = 0, (Gni)/2
do k = 1, Nkl
*JJ do jw = 1, (Maxy-Miny+1)
do jw = 1, njl
fwft_8(jw,k,2*i+1) = F_pri_8 * fwft_8(jw,k,2*i+1)
fwft_8(jw,k,2*i+2) = F_pri_8 * fwft_8(jw,k,2*i+2)
enddo
enddo
enddo
!$omp enddo
!$omp do
do k = 1, Nkl
do j = 1, (Maxy-Miny+1)
fwft_8(j,k,Gni+2) = ZERO_8
fwft_8(j,k,2) = fwft_8(j,k,1)
enddo
enddo
!$omp enddo
!$omp single
*
call rpn_comm_transpose (fwft_8(Miny,1,2),Miny,Maxy,nx3,
$ (Maxx1-Minx1+1),minx2, maxx2,gni,fdg2_8,2,2)
!$omp end single
*
* cote droit
*
!$omp do
do j = 1, nx3
do o1= 1, F_pwr
do i = minx2, maxx2
do k = minx1, maxx1
sol_8(k,i,o1,j) = ZERO_8
dn3_8(k,i,o1,j) = ZERO_8
enddo
enddo
enddo
do i = 1, nx2
do k = 1, nx1
dn3_8(k,i,1,j)= F_opsyp0_8(nx3+j)*fdg2_8(k,i,j)
enddo
enddo
enddo
!$omp enddo
*
* resolution du systeme blok-tridiagonal
*
* aller
!$omp do
do o1= 1,F_pwr
do i = 1, nx2
do k= 1, nx1
sol_8(k,i,o1,1) = dn3_8(k,i,o1,1)
enddo
enddo
enddo
!$omp enddo
*
!$omp do
do kkii = 1, Ptopo_npeOpenMP
ki0 = minx2 + kilon*(kkii-1)
kin = min(ki0+kilon-1, maxx2)
pi0 = 1+ kilon*(kkii-1)
pin = min(pi0+kilon-1, nx2)
do j = 2, nx3
do o1= 1, F_pwr
do o2= 1, F_pwr
do i = ki0, kin
do k = 1, nx1
sol_8(k,i,o1,j)= sol_8(k,i,o1,j)
$ + F_a_8(o1,o2,i,j)*sol_8(k,i,o2,j-1)
enddo
enddo
enddo
enddo
do o1= 1,F_pwr
do i = pi0, pin
do k= 1, nx1
sol_8(k,i,o1,j) = dn3_8(k,i,o1,j) - sol_8(k,i,o1,j)
enddo
enddo
enddo
enddo
enddo
!$omp enddo
*
* scale le cote droit pour retour
*
!$omp do
do j = 1, nx3
do o1= 1, F_pwr
do i = minx2, maxx2
do k = minx1, maxx1
dn3_8(k,i,o1,j) = 0.0
enddo
enddo
enddo
do o2=1,F_pwr
do o1=1,F_pwr
do i= minx2,maxx2
do k= minx1,maxx1
dn3_8(k,i,o1,j)= dn3_8(k,i,o1,j)
$ + F_deltai_8(o1,o2,i,j)*sol_8(k,i,o2,j)
enddo
enddo
enddo
enddo
enddo
!$omp enddo
*
* retour
*
!$omp do
do j = 1, nx3-1
do o1= 1, F_pwr
do i = 1, nx2
do k = 1, nx1
sol_8(k,i,o1,j)=0.0
enddo
enddo
enddo
enddo
!$omp enddo
!$omp do
do o1= 1, F_pwr
do i = 1, nx2
do k = 1, nx1
sol_8(k,i,o1,nx3)=dn3_8(k,i,o1,nx3)
enddo
enddo
enddo
!$omp enddo
*
!$omp do
do kkii = 1, Ptopo_npeOpenMP
ki0 = minx2 + kilon*(kkii-1)
kin = min(ki0+kilon-1, maxx2)
pi0 = 1+ kilon*(kkii-1)
pin = min(pi0+kilon-1, nx2)
do j = nx3-1, 1, -1
do o2= 1, F_pwr
do o1= 1, F_pwr
do i = ki0, kin
do k = minx1, maxx1
sol_8(k,i,o1,j)= sol_8(k,i,o1,j)
$ + F_c_8(o1,o2,i,j)*sol_8(k,i,o2,j+1)
enddo
enddo
enddo
enddo
do o1= 1, F_pwr
do i = pi0, pin
do k = 1, nx1
sol_8(k,i,o1,j)=dn3_8(k,i,o1,j)-sol_8(k,i,o1,j)
enddo
enddo
enddo
enddo
enddo
!$omp enddo
*
!$omp do
do j = 1, nx3
do i = 1, nx2
do k = 1, nx1
fdg2_8(k,i,j)=sol_8(k,i,F_pwr,j)
enddo
enddo
enddo
!$omp enddo
*
* inverse projection ( r = x * w )
*
!$omp single
call rpn_comm_transpose
$ ( fdg1_8 , Miny, Maxy, nx3, (Maxx1-Minx1+1),
$ minx2, maxx2,gni, fdg2_8,- 2, 2 )
*
!$omp end single
!$omp do
do i = 1, Gni
do k = 1, nkl
do j=1,njl
fwft_8(j,k,i+1) = fdg1_8(j,k,i)
enddo
enddo
enddo
!$omp enddo
!$omp do
do k = 1, nkl
do j = 1, (Maxy-Miny+1)
fwft_8(j,k,1) = fwft_8(j,k,2)
fwft_8(j,k,2) = ZERO_8
fwft_8(j,k,Gni+2) = ZERO_8
enddo
enddo
!$omp enddo
*
!$omp do
do k=1,Nkl
call ffft8 (fwft_8(Miny,k,1),(Maxy-Miny+1)*(maxx1-minx1+1),1,
% (Maxy-Miny+1), +1 )
enddo
!$omp enddo
!$omp end parallel
call rpn_comm_transpose48 ( F_sol, Minx, Maxx, Gni,1, (Maxy-Miny+1),
% (Maxy-Miny+1),minx1, maxx1, Gnk, fwft_8, -1,
% 1.d0,0.d0)
*
* __________________________________________________________________
*
return
end